Chapter 1

GENERAL INTRODUCTION
AND
REVIEW OF HISTORICAL DEVELOPMENT

1.1: On Generalizations of Graphs

At present, nobody will deny the importance and overwhelming craze for

application of graph theory to various aspects of theoretical and practical fields of

activities. However, graphs only describe some binary relations and are not always

sufficient for modelling problems or data which involve relations of order higher than

binary. In order to tackle additional combinatorial problems involving any higher order

relation, it was therefore natural to model and generalize the concept of graph in such a

way that it allows applications to wider spectrum of real life situation. The first attempt

in this regard was made by C. Berge [5] who in 1960 shaped the idea of what we now

call a “Hypergraph”. Primary motivation behind this idea rested on the observation that

an edge of a graph G can be interpreted as a 2-element subset of the vertex set V of G

and therefore it can be naturally extended to an edge interpreted as a subset consisting

1
 2

of more than two elements of V. For instance, people going to a market can have some

sort of binary relation like one buying and another selling leading to the formation of a

graph in its ordinary sense; but there are occasions in which mutual transactions are

made with the help of a broker and this situation cannot be expressed by our so called

graphs. In fact, seller, buyer and the broker form a 3-element subset of the set of people

in the market so that the existing concept of edge with 2-element set can be extended to

an edge with 3-element set leading to emergence of the concept of a hyper edge of a

hypergraph. In the literature, plenty of beautiful examples of hypergraphs exist, which

are not strongly connected to the present investigation and therefore we refrain from

making any detailed reference of them in present work.

The second attempt to generalize the concept of graph was concentrated upon

the fact that a graph G consists of a set V of elements called vertices together with a

prescribed set X of unordered 2-tuple {u, v} called an edge which can be pictorially

denoted by joining u and v by a line where u, v ϵ V . Capitalizing on this aspect of edge

as a line, its 2-tuple set notation {u, v} can be modified as (u, v) where both (u, v) and

(v, u) are considered to represent the same edge. In other words, the edges (u, v) and

(u ′, v′) represent one and the same edge if and only if either u = u ′ and v = v ′ or,

u = v ′ and v = u ′ . With this, the path was opened up for another generalization of the

edge of a graph and such a generalization was successfully initiated by a famous Indian

graph theorist named Sampathkumarachar [19] who proposed an edge with more than

two vertices viz., (u1 , u 2 , u 3 ,..., u n ) where the edges (u1 , u 2 , u 3 ,..., u m ) and (v1 , v 2 ,..., v n )

are considered to represent the same edge if and only if m = n and either u i = vi for

1≤ i ≤ n or u i = v n −i +1 for 1≤ i ≤ n. However, in order to retain the original structure of

graph, where two distinct edges intersect in at most at one vertex only, it became

necessary for the new edges to satisfy the same condition that two distinct edges in the

new structure intersect in at most at one vertex.

Thus, in spite of the fact that both hypergraph and semigraph generalize an edge

to contain more than two vertices, the particular order imposed upon vertices of the

edge of a semigraph makes the semigraph more akin to a graph than a hypergraph.

Arrangement of vertices of an edge in a particular order in a semigraph has few special

advantages over its counterpart viz., hypergraph as mentioned in [19]. These are:

(i) The important concept of planarity of ordinary graph theory can be

translated in a straightforward manner to semigraphs while the same

cannot be realized for hypergraphs [19], [46] and [55].

(ii) The concept of Eulerian and Hamiltonian paths of ordinary graphs can be

easily introduced in semigraphs leading to the definition of Eulerian and

Hamiltonian graphs which is not possible in hypergraphs.

(iii) Since each edge in a semigraph looks like an edge in a graph, one can

easily give a direction to each edge and obtain a structure called directed

semigraph, analogue of which is not possible in hypergraphs.

While the justifications as enumerated in the preceding lines were sufficient for

the new generalization of graph called semigraph its originator E. Sampathkumar [19]

also recorded the following observations:

(i)′ In graph, any number of mutually non-adjacent points may be adjacent to

same point while similar phenomenon cannot be found for edges.

(ii)′ Analogous to the concept of block graph B(G) of a graph G, where every

vertex represents a block of G and two vertices are adjacent in B(G) if

and only if the corresponding blocks in G are adjacent, we do not have a

concept of graph where each edge may represent a block of G.

(iii)′ Analogous to the concept of line graph of a graph G, we do not have a

concept of a point graph where each edge represents a vertex of G.

This new generalization of graph originally named as graphoid was renamed

later on as semigraph by B. D. Acharya in 1994 and it was structurally designed so as to

apply to problems as demanded by situations in real life that involve relations more than

binary ones.

In this context interestingly, one could also note that any property enjoyed by

vertices as mentioned in (i)′ is also enjoyed by edges in the case of semigraph.

One concrete example of application of semigraph would perhaps clear all

doubts as to why a generalization like semigraph should be created in addition to

hypergraph. While encountering with problem for DNA splicing S. Jeyabharathi et al.

[53] and [54], found that semigraph is more effective tool than graphs for describing

DNA splicing, thereby establishing the requirement for creation of such a

generalization.
 5

1.2: Literary Review

After being confirmed that the new generalization was totally natural and worthy

for applications to wider variety of situations, E. Sampathkumar [19] decided to build a

structural foundation for it. His immediate task was to examine and see how does the

new idea work and what impact does it have on the existing terminology of graph

theory. The immediate impact was overwhelming. With recognition of edge through

more than two vertices in case of semigraph the concepts of adjacency of vertices and

edges and also the degree of vertices had to be extended thereby creating adjacency, end

vertex adjacency, consecutive adjacency, degree, edge degree, adjacent degree and

consecutive adjacent degree (all mentioned in Section 2.1 of Chapter 2). With modified

edge concept there were new entrants like subedge and partial edge of a semigraph

followed by other concepts like edge complete semigraphs, edge bipartite semigraphs

and edge cliques which are extensions of similar ideas in ordinary graphs. On the basis

of subedges and partial edges, Sampathkumar introduced the concept of subsemigraphs

and partial subsemigraphs. Similarly, two types of paths and cycles called s-path (strong

path), w-path (weak path), s-cycle (strong cycle) and w-cycle (weak cycle). The concept

of independence of vertices had its three new versions viz., independence,

e-independence and strongly independence in the context of semigraph and on the basis

of these independence concepts three types of bipartite semigraphs were defined namely

bipartite, e-bipartite and strongly bipartite semigraphs. E. Sampathkumar derived three

types of new graphs from a semigraph which played significant role in the theory of

semigraph. The three graphs associated with a semigraph are End vertex graph,
 6

Adjacency graph and Consecutive adjacency graph. The cut vertex, bridge and block are

also defined for a semigraph with a characterization theorem for a block of a semigraph.

In the same work [19], E. Sampathkumar introduced a concept analogous to

trees in graphs called dendroids. The idea of eccentricity, radius, diameter and center are

also discussed in the work with necessary modifications. Due attention was also paid to

the concepts of cycle vector, cycle space, cycle basis, cycle rank in the context of

semigraphs.

While considering mappings of semigraphs four different types of isomorphisms

viz., isomorphism, end vertex isomorphism, edge isomorphism and adjacency

isomorphism emerged and their interconnections are examined. Particularly, relations

among complements of different isomorphisms of semigraphs are examined in this

transitory work.

The next issue to be taken up in this line of action was the covering of graph and

E. Sampathkumar was successful in defining two types of coverings for each of vertex

covering and edge covering for semigraphs. These are named as vertex covering and

e-vertex covering for the first and edge covering and e-edge covering for the second of

the two types followed by results showing their necessary characterizations.

As is evident from what has been discussed already, graph theoretic terms and

definitions seemed rallying one after another to be accommodated in semigraph setting

either for their extensions or new interpretations. With generalized graph structure
 7

allowing unusually long list of terms and definitions which were algebraically screened

with positive outcomes it was natural to eye upon its topological aspects and,

expectations were fruitfully rewarded!

Naturally, new target was connectivity. Looking at existence of two types of

paths viz., strong path (s-path) and weak path (w-path) in a semigraph E. Sampathkumar

introduced two types of connectivity (path connectivity) viz., s-connectivity and

w-connectivity for semigraph. Also, depending upon subedge (fs-edge) and partial edge

(fp-edge) of any edge another two types of connectivity (edge connectivity) viz., s-edge

connectivity and w-edge connectivity were introduced. Two Menger-type theorems

concerning each of the (path) connectivity and edge connectivity were obtained in this

context.

The concept of planarity in semigraph was rather a straightforward

generalization of planarity in graph and accordingly the Euler’s polyhedral formula was

easily generalized for planar semigraphs. It was interesting to note that Kuratowski’s

theorem for graph could be used to characterize planar semigraphs. Some minimal non-

planar semigraphs for five and six vertices were also obtained.

Then E. Sampathkumar [19] introduced the concept of dual of a semigraph for

which every vertex has edge degree at least two. It was noted that dual of a Semigraph

was not unique. He examined semigraphs having unique duals and those having self

duals. Some necessary conditions for a semigraph to be self-dual were obtained.

Eulerian and Hamiltonian graphs are amongst the most interesting and much

talked about topics of graph theory from its inception studied under traversability. The

way graph is generalized to define hypergraph does not provide necessary structure for

introducing the concept of traversability in it. However, E. Sampathkumar was

successful in introducing it in semigraph. He started with the idea of adjacency disjoint

subedges of an edge and defined three types of partitions of an edge viz., s-partition,

p-partition and 2-partition which helped him introduce as many as four types of

Eulerian semigraph viz., s-Eulerian semigraph, p-Eulerian semigraph, 2-Eulerian

semigraph and f-Eulerian semigraph. Then he went on to establish some results showing

interdependence of these semigraphs. Particularly, he deduced that (i) an f-Eulerian

semigraph is both s-Eulerian and p-Eulerian. (ii) A 2-Eulerian semigraph is s-Eulerian.

He also derived two necessary and sufficient conditions for a semigraph to be

p-Eulerian.

Based on the definition of Hamiltonian cycles for semigraph, E. Sampathkumar

first proposed the definition of Hamiltonian semigraph. Then, introducing the notions of

s-cycle and w-cycle he proposed the definitions of s-Hamiltonian and w-Hamiltonian

semigraphs respectively. It was noted that any Hamiltonian semigraph is

w-Hamiltonian. Also an s-Hamiltonian semigraph is w-Hamiltonian but not conversely.

He obtained a good number of substantial results in this regard. Particularly, he

established the semigraph versions of some classical graph theory results due to Dirac

[24], Ore [43], Bondy and Chvatal [26].

Like other graph theoretic ideas analogue of line graph was easily

accommodated in semigraph resulting in two other semigraphs of this type viz., line

semigraph and point semigraph. While line semigraph needed edge labeling in one hand

and point semigraph that of adjacency of points on the other, both of the two

semigraphs, in addition, utilized the concept of isomorphism of semigraphs. Also there

being different types of adjacency relations in respect of points and edges for

semigraphs a volley of new definitions of line and point semigraphs resulted thereafter.

A rich collection of results both classical and new followed in this context [19].

The famous classical problems of ‘Map Coloring’ popularly known as the four

color theorem and five color theorem which are discussed in graph theory as coloring of

graph have been examined in semigraphs also [19]. Three types of vertex coloring

named as coloring, e-coloring and strong coloring were defined and their corresponding

chromatic numbers were introduced. A result determining relationship between these

chromatic numbers was obtained. Some other results determining chromatic numbers of

different types of semigraphs were also discussed. A problem concerning chromatic

number of a complete semigraph was raised and its solution is yet to be realized.

Depending upon an edge being adjacent or ee-adjacent it was possible to define

two different types of edge coloring viz., edge coloring and e-edge coloring.

Accordingly, the concepts of edge chromatic number and e-edge chromatic number

could be defined and discussed. Few results involving bounds for each of the edge

chromatic number and e-edge chromatic number which coincide with chromatic index

for graphs were obtained.

Semigraphs like graphs can be represented by matrices which make the study of

semigraphs possible in certain of its aspects. As in other cases here also the original

matrix representation of graph gives way to new incarnations in semigraph setting

which revealed in the forms of the adjacency matrix, the incidence matrix, the

consecutive adjacency matrix and the 3-matrix of a semigraph. It was found that the

incidence matrix together with the consecutive adjacency matrix determines a

semigraph uniquely. Also, the 3-matrix of a semigraph G determines the semigraph G

uniquely.

The generalization process of graph in the form of semigraph due to

E. Sampathkumar culminated with an introduction and discussion on directed

semigraph in his innovative project “Semigraphs and Their Applications” [19]. The

necessary pre-requisites being already available and subsidiary related ideas being

already developed considerably, it simply required the attachment of orientation to the

edges of the semigraph to make it into a disemigraph. Analogous to subedge, partial

subedge, subsemigraph etc., subarc, partial arc, subdisemigraph are introduced for

directed semigraph (or disemigraph) so that walks, trails, paths, cycles etc., can be

defined for it. A result of directed graph involving in-degree and out-degree was

generalized to disemigraph. Some results on connected disemigraph, component of

disemigraph and eulerian disemigraph were also derived.

Within a very short period after E. Sampathkumar contributed the new

generalization of graph to the world of Mathematics in 2000, Indian graph theorists

mostly from southern and western parts of the country took up keen interest to this new

area and they began attempting to complete the untouched and unfinished works of its

originator. In 2003, S.S. Kamath and R.S. Bhat [57] took up one of the very popular and

important concepts viz., domination of graph into semigraph. In this paper, they

introduced three types of dominating sets for semigraphs viz., adjacent dominating set,

end vertex dominating set and consecutive adjacent dominating set along with their

corresponding domination numbers. They obtained these domination parameters along

with their bounds for various semigraphs.

In the same year, S. S. Kamath along with Saroja R. Hebbar extended the

concepts of adjacent dominating vertices and adjacent dominating sets of semigraphs

and proposed strongly (weakly) adjacent dominating vertices, strongly (weakly)

dominating sets, full sets and domination balance. In another paper they forwarded the

concept of domination critical semigraphs [58], [59].

In the year 2004, B. Y. Bam [3] of Pune University carried out the research on

semigraph to give a solution of the e-semigraphical problem raised by E. Sampathkumar

[19]. To describe the nature of vertex in an edge of a semigraph, B. Y. Bam introduced

a new concept of degree called as me-degree or (m, e)-degree. Using this new concept

he obtained some necessary conditions for a (m, e)-degree sequence to be me-

semigraphical and raised an open problem on characterization of me-semigraphical

sequences. In both of [3] and [42], B. Y. Bam considered the issue of line semigraph

and presented a definition of line semigraph different from that given by

E. Sampathkumar. A characterization of line semigraphs was obtained in this regard.

The work on matrix representations of Semigraphs viz., Adjacency matrix,

Incidence matrix, Consecutive adjacency matrix and the 3-matrix of semigraphs as

reviewed earlier was published in 2007 by E. Sampathkumar and L. Pushpalata [20].

Further development was brought to the theory of semigraphs through a series

of four collaborated papers by K. Kayathri, Mary Sunithi Vijayan and S. Pethanachi

Selvam, first three of which were published in 2007 and the last in 2010. The first paper

authored by K. Kayathri and Mary Sunithi Vijayan [31] dealt with the problem of

coloring of complete semigraph and obtained chromatic numbers for some special types

of semigraphs. The second and third papers in the series authored by K. Kayathri and S.

Pethanachi Selvam ([32], [33]), dealt with edge completeness of (p, 2) and (p, 3)

semigraphs. The last mentioned authors collaborated again in 2010 and introduced the

concept of enumeration of edge complete (p, 3) semigraph [34] in which they found 20

categories of edge complete (p, 3) semigraphs and proved that no two of these 20

categories of semigraphs were isomorphic. They also worked on enumeration of non

isomorphic edge complete (p, 2) semigraphs.

The study of N. S. Bhave and B. Y. Bam [3] on (m, e)-degree of vertices etc.,

was revived again in 2009 by S. Gomathi, R. Sundareswaran and V. Swaminathan [52]

and they extended it to the concept of (m, e)-strong dominating set, (m, e)-dominating

set, (m, e)-domination number along with a characterization theorem of (m, e)-strong

dominating set.
 13

The works outlined in the preceding paragraphs give sufficient testimony about

the strength of semigraph as a mathematical model and its applicability to various

aspects of formal structures. However, no model, however pure and perfect may be in

form it becomes useful only when it is tested by physical application. Semigraph

overcame an important physical test when the concept of semigraph was used for DNA

splicing. In this context we refer to the works of S. Jeyabharathi, J. Padmashree, S. S.

Selvi and K. Thiagarajan [53]. Then K. Thiagarajan, S. Jeyabharathi and L. Pushpalatha

[35] correlated the concept of the graph splicing with semigraphs and obtained a

characterization of graph splicing languages in terms of semigraph to show some

splicing graph properties. In 2012 (December), S. Jeyabharathi, M. Angayarkanni, S.

Sinthanai Selvi and R. Anusha [54], observed that, at the time of splicing DNA molecule

holds the characterization of semigraph, which is more powerful than graph model. In

this paper they also discussed the properties of the odd and even cut bipartite semigraph

structures of double stranded DNA molecules and characterized the number of

independent sets and languages produced by them.

Based on various types of domination concepts as presented in the works of

S. Gomathi, R. Sundareswaran and V. Swaminathan [52], last year, i.e. in 2012, Y. B.

Venkatakrishnan and V. Swaminathan [64] introduced some bipartite graphs

corresponding to a semigraph. They obtained the equality of domination parameters in

bipartite graphs constructed with the domination and total domination parameters of

adjacency and consecutive adjacency graphs associated with a semigraph. They also

introduced the domination and independence parameters for bipartite semigraph and

defined Xa-chromatic number, Xa-hyperindependent number, Xa-irredundant number

and Xa-dominating sequence chain. Recently, they worked out results on the concept of

bipartite semigraphs and successfully introduced hyper domination of a vertex, hyper

domination of a set, hyper domination number of a set, hyper independent set and hyper

irredundant set in [65]. Some inequalities involving dominating parameters and

irredundant parameters were also obtained in this regard.

Encouraged by new developments in various directions of semigraph structure

new faces are attracted into its fold day by day. In 2011, a collaborated paper on

Directed Semigraph (Disemigraph) was published by C. V. R. Harinarayanan, S. P.

Subbiah, C. Y. Ponnappan, R. Sundereswaran, V. Swaminathan [10]. They defined a

topology on the vertex set of a disemigraph in such a way that there exists one to one

correspondence between the set of all topologies on the vertex set having the property of

completely additive closure and the set of all equivalence classes of disemigraph

defined on the vertex set.

Recently (2012), P. Das and Hamida Aktra Hoque [45] attempted to obtain a

tournament like structure in disemigraph. Their venture was successful after a

modification in the corresponding definition of tournament in digraph. A

characterization of tournament in disemigraph was also established.

As per available information regarding the latest position of publications on

relevant topics in semigraph the last paper is due to C. M. Deshpande and Y. S.

Gaidhani [7] (2012, July) and it gives a new definition for the adjacency matrix of a

semigraph. They also introduce the notion of spectrum of a semigraph and study some
 15

of its spectral properties. Moreover, they obtain a necessary and sufficient condition for

a matrix to be semigraphical.

1.3: Organization of the Thesis

The framework of this thesis is to carry forward the structural features of the

theory of semigraph as conceived by E. Sampathkumar as a generalization of graph

theory. In spite of many researchers are engaged in the study of various aspects of the

semigraph immediately after its inception, there is a lack of cohesion as well as co-

ordination in these endeavours and a lot of works in the direction of algebraical and

topological structures of semigraphs still remains to be done. This thesis is a humble

attempt to supplement few important areas of semigraph by introducing new ideas in

conformity with the spirit of the subject. Mainly here, the concepts of factorization,

matching and few topological invariants like genus, thickness, coarseness and crossing

number of graphs are generalized into semigraphs. Some new concepts like maximum

vertex-saturated, maximum edge-saturated and optimum matching concerning the

concept of matching and generator of a face in connection with genus in semigraphs are

also considered here. Total adjacent domination and arboricity are also two other

concepts introduced here. Some vital conditions of 1-factorization of graph not held in

case of semigraphs are demonstrated with the help of concrete examples. The

organization of the thesis is as follows.

The thesis consists of six chapters. In the first chapter a general introduction

regarding motivation of the problem followed by a discussion on its recent development

through works of almost all well-known researchers working on semigraph after

E. Sampathkumar up to date has been incorporated. Indeed, this literature survey has

clearly justified the relevance and importance of the problem under consideration. A

chapter-wise introduction in brief is also presented in this chapter so that reader can

make some prior idea about the content of the thesis.

The second chapter introduces the reader with a collection of notations and

ideas in the form of definitions, axioms and results which can be used as ready reference

for easy understanding of the subsequent topics. The collection of notations and ideas in

semigraphs like graphs is usually very rich and though many of them are yet to be

recognized as standard however, they have found place in the recent works on

semigraphs and therefore, these are simply borrowed and reproduced without any

change here.

The Third chapter deals with the concept of factorization in semigraph. Based on

different types of degree of vertices in semigraph particularly, the degree and edge

degree, two types of factorization viz., 1-factorization and 1e − factorization are

introduced here. Encouraged by the study of Tutte’s [60] significant result on

characterization of a graph to have 1-factorization similar result for semigraph was

attempted. However, no such characterization is obtained to give any guaranty in such a

case which has been demonstrated with the help of example. Of course, similar to

relation between semigraph and classification problem due to Cariolaro [14] some
 17

results are established. Another graph factorization problem called arboricity is also

introduced for semigraph and a basic result of it has been established.

The fourth chapter is devoted to establishing the concept of matching in

semigraph. On the basis of the C. Berge’s [5], [6] characterization theorem for matching

in graphs as well as in hypergraphs, a similar result has been established for semigraphs.

A result analogous to König’s [16] theorem in graph theory is also obtained. Some new

concepts namely, maximal vertex-saturated matching, minimal edge-saturated matching

and optimal matching along with their parameters known as power of maximal

vertex-saturated p ( M mvs ) and power of minimal edge-saturated p( M mes ) are introduced

and demonstrated with relevant examples here. This chapter also deals with the total

adjacent domination concept for semigraph along with some results and a result

showing relation between total adjacent domination number and edge independence

number has been established. The chapter concludes by establishing a relation between

total adjacent domination number and power of minimal edge-saturated matching for

semigraph.

In the fifth chapter an important topological property called genus of semigraph

is examined. The similar concept in graph theory can be traced back to “Euler

Polyhedron Formula” after which the concept of planar graphs was developed and the

parameter called genus for non-planar graph was originated. Heawood’s conjecture [49]

can also be recalled in this respect. However, the concept of genus does not

automatically enter into semigraph. A new concept namely, ‘the generator of a face’ is

needed to introduce the definition of genus for a semigraph and this has been
 18

successfully done for a compact orientable 2-manifold surface. In a sense, this result

generalizes the Euler’s formula for graph into the semigraph setting. Another important

result on lower bound for genus for each of a connected semigraph with each face

having triangle and a connected semigraph with each face having no triangle has been

obtained. A result on lower bound for a complete semigraph has also been obtained

here. Most of these results are verified by illustration of concrete examples.

The sixth, also the last major chapter deals with some other topological

parameters viz., thickness, coarseness and crossing number for semigraphs. The origin

of the thickness problem can be traced back to 1961appearing in a ‘Research Problem’

by Harary [21] in sequel to an attempt to prove or disprove whether “For any graph G

with 9 points, G or its complementary graph is non-planar”. Tutte [63] introduced the

concept of thickness to generalize this problem and Beineke and Harary [37] found

thickness for certain complete graphs. The concept of coarseness on non-planar graphs

was introduced by P. Erdös (verbal communication) and, Beineke and Guy [39]

investigated coarseness of complete graphs. Turán's brick factory problem [50] was the

introducer of the crossing number problem of non-planar graphs. Guy [51] described

some results on this concept. Beineke and Chartrand [40] found some bounds of

coarseness of complete graphs. The generalization of these parameters namely thickness

θ (G ) , coarseness ζ (G) and crossing numberν (G) involving a non planar semigraph G

is studied here. Mainly, the problem of determination of lower bound of thickness for

any semigraph, complete semigraph and connected semigraph containing no any

triangle is dealt with separately. Based on the result obtained by Beineke and Chartrand

[40] in graphs, the upper bound of coarseness for each of semigraph and complete
 19

semigraph is established separately. Also, three results, one each for determining lower

bound of the crossing number of a semigraph, a complete semigraph and a semigraph

without any triangle are obtained.

The thesis is concluded with a brief comment at the end.

*****